Abstract
Exposure to simple patterns changes neural responses in early visual cortex, an effect known as contrast adaptation. However, when these changed signals propagate to later visual areas they may produce inaccurate responses there, as illustrated by illusions like the tilt aftereffect. Later regions are themselves capable of adapting, however, as in face adaptation. Here we tested whether such higher-level adaptation can normalize, and possibly correct, response to adapted input from earlier areas. Specifically, we tested whether distortions in faces produced by contrast adaptation can be affected by higher-level adaptation, causing the distorted faces to appear more normal. Subjects adapted for 2 min to contrast-reversing checkerboard patches presented near fixation; this contrast adaptation displaced the eyes of a normal face presented at fixation, causing them to appear shifted outward. Subjects viewed 300 msec presentations of these distorted faces in alternation with 6 sec of top-up checkerboard adaptation. Every six repetitions of this face-top-up cycle, subjects judged a test face image presented above fixation. Pilot data showed that the checkerboard adaptation did not transfer to this location, confirming its origin in early visual cortex. Participants judged whether the test face's eyes appeared to be closer together or further apart than "normal". Interocular distance of the test face was controlled by a staircase procedure to estimate the distance that appeared normal. Our results showed that after adaptation, subjects' (n=7) estimates of normal eye position in the test face were shifted significantly (t(6)=-2.23, p< 0.05), in the direction of the distortion produced by the contrast adaptation (outward). Thus, high-level adaptation caused the distorted faces to appear more normal. This pattern suggests that adaptation in early visual areas can propagate up the visual stream and generate adaptation in later regions. Such "meta-adaptation" could possibly correct for coding errors generated by adaptation in early visual areas.
Meeting abstract presented at VSS 2018